Triarylmethyl cation redox mediators enhance Li–O2 battery discharge capacities

A major impediment to Li–O2 battery commercialization is the low discharge capacities resulting from electronically insulating Li2O2 film growth on carbon electrodes. Redox mediation offers an effective strategy to drive oxygen chemistry into solution, avoiding surface-mediated Li2O2 film growth and extending discharge lifetimes. As such, the exploration of diverse redox mediator classes can aid the development of molecular design criteria. Here we report a class of triarylmethyl cations that are effective at enhancing discharge capacities up to 35-fold. Surprisingly, we observe that redox mediators with more positive reduction potentials lead to larger discharge capacities because of their improved ability to suppress the surface-mediated reduction pathway. This result provides important structure–property relationships for future improvements in redox-mediated O2/Li2O2 discharge capacities. Furthermore, we applied a chronopotentiometry model to investigate the zones of redox mediator standard reduction potentials and the concentrations needed to achieve efficient redox mediation at a given current density. We expect this analysis to guide future redox mediator exploration. Although Li–O2 batteries offer high theoretical capacities, redox mediators are necessary to control intermediate reaction kinetics and to limit electrode passivation. Now it has been shown that a family of triarylmethyl cations can rival top-performing quinone-based redox mediators. Cations with sluggish catalytic rates were found to suppress surface-mediated O2 reduction and achieve higher capacitances.

A major impediment to Li-O 2 battery commercialization is the low discharge capacities resulting from electronically insulating Li 2 O 2 film growth on carbon electrodes. Redox mediation offers an effective strategy to drive oxygen chemistry into solution, avoiding surface-mediated Li 2 O 2 film growth and extending discharge lifetimes. As such, the exploration of diverse redox mediator classes can aid the development of molecular design criteria. Here we report a class of triarylmethyl cations that are effective at enhancing discharge capacities up to 35-fold. Surprisingly, we observe that redox mediators with more positive reduction potentials lead to larger discharge capacities because of their improved ability to suppress the surface-mediated reduction pathway. This result provides important structure-property relationships for future improvements in redox-mediated O 2 /Li 2 O 2 discharge capacities. Furthermore, we applied a chronopotentiometry model to investigate the zones of redox mediator standard reduction potentials and the concentrations needed to achieve efficient redox mediation at a given current density. We expect this analysis to guide future redox mediator exploration.
The Li-oxygen battery exhibits a tremendously high theoretical specific energy and is receiving interest as a next-generation energy system. In a typical cell, O 2 is reduced through a series of Li + -coupled electron-transfer steps or disproportionation to grow Li 2 O 2 (refs. 1,2). Although the Li-O 2 battery promises large energy density and capacities, the extent to which it can deliver on this promise is contingent upon the mechanisms by which solid Li 2 O 2 growth is mediated. After its initial reduction at the carbon cathode, subsequent steps involving LiO 2 intermediates may proceed in different locations. For example, in one case, LiO 2 may stay on the electrode surface and be reduced to Li 2 O 2 , generating a thin layer of Li 2 O 2 , which electronically passivates the electrode, resulting in premature cell death 1,3 . Alternatively, LiO 2 may dissolve into the electrolyte and, through disproportionation, generate Li 2 O 2 away from the electrode, delivering large battery capacities. With normal Li-O 2 electrolytes (additive-free, non-polar organic solvents), the solubility of reduced oxygen intermediates is low, and the electrode-mediated process dominates. Therefore, it is paramount to control the oxygen reduction reaction (ORR) pathway and drive it into solution.
To boost the solution-based mechanism of Li 2 O 2 formation in Li-O 2 batteries, polar species 4-7 and redox-active compounds were tested as additives [8][9][10][11][12][13][14][15][16] . These additives move the Li 2 O 2 formation site from the electrode surface into solution, but their mechanism of action varies. Polar species increase the solubility of Li + or LiO 2 intermediates but are understood to negatively impact long-term cyclability 7 . A more effective approach is redox mediation, wherein small molecules shuttle electrons from the electrode to O 2 . Quinone-based redox mediators promote solution discharge via an inner-sphere mechanism involving an oxygen-quinone intermediate formed by reaction between reduced quinone and O 2 (refs. [8][9][10]12,13). Li-O 2 batteries with quinone-based additives have achieved up to 100-fold capacity improvement 9 and exhibited toroidal Li 2 O 2 formation on the electrode, indicating solution-phase growth 17 . Outer-sphere redox mediation has also been explored with viologen derivatives, which serve as electron shuttles that facilitate superoxide formation away from the electrode 14,15 . Additionally, non-redox-active catalysts, based on the biomimetic catalytic disproportionation of superoxide into O 2 and peroxide 16 , have recently been investigated.
Once discharged, recharging the battery presents itself as another performance-limiting hurdle for Li-O 2 batteries. Large overpotentials are required to oxidize Li 2 O 2 deposits. At such oxidizing potentials, cell components such as the solvent and carbon electrode are susceptible to degradation. Redox mediation is again an effective strategy to address these concerns. Here, redox mediation proceeds through oxidation of the mediator at the electrode at low overpotentials. Oxidized mediators deliver an electron hole needed for Li 2 O 2 oxidation. Several inorganic anions have been introduced as charging redox mediators (I − , Br − and NO 3 − ) [18][19][20] . However, the inability to tune oxidation potentials and the negative effects on battery efficiency have led researchers to explore organic compounds for charge redox mediation 11,18,19,[21][22][23][24][25] . The nitroxyl-based (2,2,6,6-tetramethylpiperidin-1-yl)oxyl (TEMPO) demonstrated the best performance and stability. Indeed, when joint 2,5-di-tert-butyl-1,4-benzoquinone (DBBQ)/TEMPO discharge/charge redox mediators are used, long-term stability has been achieved, suggesting that joint redox mediators are a viable path towards a highly efficient and long-cycling Li-O 2 battery 26 .

Inner-sphere pathway
Outer-sphere pathway Step 1 Step 2 Step 3 Article https://doi.org/10.1038/s41557-023-01268-0 ( Fig. 1b). Some R • in Fig. 1a have been shown to react rapidly with O 2 to form RO-OR 36 , which prompted us to investigate this mechanistic pathway. Catalytic cycle closure in the inner-sphere pathway is achieved by the formation of Li 2 O 2 through reaction of RO-OR with Li + .

Computational screening
Thermodynamic evaluations of the inner-and outer-sphere mechanisms were performed at the DFT level of theory by calculating ΔG 0 values for the formation of R • (ΔG 0 1 ; Fig. 1b, step 1), the reaction between R • and O 2 (ΔG 0 2 , step 2), and the reaction between RO-OR and Li + (ΔG 0 3 , step 3). Computational details, including free-energy calculations and an explanation of the screening criteria, are provided in Supplementary Section 2. Using calculated E 0 (R + /R • ) and ΔG 0 values, we constructed a flowchart that predicts the activity of proposed cations as inner-or outer-sphere redox mediators (Supplementary Scheme 1).
Step 1 compares E 0 (R + /R • ) with an established electrochemical window for oxygen catalysis (2.96-2.  Table 6). As expected, E 0 (R + /R • ) shifted to more positive potentials with the introduction of electron-withdrawing heteroatoms (for example, oxygen or nitrogen) or substituents (for example, fluorine) and more negative potentials with electron-donating substituents (for example, methoxy). As such, compounds 3c, 3e, 3f and 5 were removed based on their E 0 (R + /R • ) values being outside the established potential window.
The  Table 6). Positive ΔG 0 2 values (1b, 1c, 1d, 1e, 2b and 4) suggest that the inner-sphere reaction with O 2 is non-spontaneous. The steric bulk imparted by substituted phenyl rings at the C9 positions of each of these compounds restricts access to the tertiary carbon-centred radical. These cations were viewed as candidates for the outer-sphere process and were evaluated using CV. For compounds with negative ΔG 0 2 values (1a, 2a, 2c, 3a, 3b and 3d), formation of RO-OR through the inner-sphere pathway is expected to be spontaneous. Here, ΔG 0 3 values reflect the feasibility of RO-OR reacting with Li + and regenerating R + . Highly positive ΔG 0 3 values suggest that 1a, 2a and 3a form stable peroxides incapable of reaction with Li + , and so they were removed as candidates. Conversely, the electron-donating substituents of 2c, 3b and 3d decrease the acidity of R + , making a less stable RO-OR and providing mild ΔG 0 3 values, making them suitable candidates for inner-sphere redox mediation. Figure 2b illustrates how calculated thermodynamic parameters affect redox mediation mechanisms. Three organic cations expected to perform redox mediation, via the inner-or outer-sphere mechanisms (compounds 1c and 2c, respectively), or be catalytically inactive (compound 2a) are shown. The diagram, presented at an applied potential of 3.0 V, shows anticipated overpotentials for each process. All three mediators exhibit positive or mildly negative ΔG 0 1 values. The positive ΔG 0 2 for 1c indicates that RO-OR is unlikely, instead favouring the outer-sphere mechanism. In contrast, the negative ΔG differentiates 2a and 2c. ΔG 0 3 is highly positive for 2a, indicating that the RO-OR is stable and will not react with Li + . Thus 2a is expected to be non-catalytic. ΔG 0 3 is only mildly positive for 2c, suggesting that it may reasonably function as an inner-sphere redox mediator. Based on these criteria, cations 1b, 1c, 1d, 1e, 2b, 2c, 3b, 3d and 4 are viewed as promising candidates for discharge redox mediation in Li-O 2 batteries via the inner-or outer-sphere mechanism (Supplementary Table 6).

Cyclic voltammetry
Redox mediation candidates were investigated with CV in Li + -containing Ar-and O 2 -saturated tetraethylene glycol dimethyl ether (TEGDME) solutions ( Fig. 3 and Supplementary Fig. 1  Article https://doi.org/10.1038/s41557-023-01268-0 a single chemically reversible R + /R • reduction was observed for all R + cations, which are plotted in order of decreasing potentials. Generally, Table 6) and their trends are rationalized using electronic effects described in the previous section. Several behaviours were observed in the presence of O 2 . In one case, the R + /R • peak of 2c (Fig. 3c) was largely unaffected by O 2 , signalling that there is no follow-up reaction between R • and O 2 on our CV timescales. The highly positive E R + /R • of 2c made R • insufficiently reducing to facilitate electron transfer to O 2 or formation of RO-OR. However, this behaviour is limited to the kinetic window of CV measurements, and the inner-sphere behaviour of 2c may operate on longer timescales.
In the case of 2a and 2b (Fig. 3a,b), cathodic R + /R • features shift to more positive potentials and anodic peaks disappear, consistent with successive electrochemical and chemical steps (EC-type mechanism), wherein R + /R • reduction is followed by reaction with O 2 , forming RO-OR through the inner-sphere pathway. The presence of an irreversible oxidation feature in the anodic scan, corresponding to RO-OR oxidation, was evidence for RO-OR formation ( Supplementary Fig. 2) 36 . Lack of catalytic current enhancement to R + /R • reduction suggests that 2a and 2b terminally formed RO-OR and no follow-up reaction with Li + occurred. As confirmation, chemically and electrochemically derived RO-OR species 2a/RO-OR and 2b/RO-OR were prepared. Both derivatives show the same UV-vis results lacking the broad absorption between 350 and 500 nm of their R + precursors (2a and 2b; Supplementary Fig. 2). Adding Li + to solutions of 2a/RO-OR and 2b/RO-OR showed no recovery of the UV-vis signatures of 2a and 2b, leading us to conclude that 2a and 2b are non-catalytic. Although 2a and 2b did not show any inner-sphere catalysis, it remains plausible that other redox mediators, with appropriate E R + /R • and that form less stable peroxides, will exhibit inner-sphere catalysis.
Compounds with the highest degree of catalytic enhancement came from the family of acridinium-based redox mediators, which were all predicted to function through the outer-sphere mechanism (1e, 1c, 1b, 1d and 4). Their CVs are shown in Fig. 3d-h and are characterized by taking 1c (Fig. 3e) as an example: in the presence of O 2 , R + /R • reduction shifts to more positive potentials, becomes chemically irreversible, and experiences a marked current enhancement. The credibility of the predicted outer-sphere mechanism was examined by a series of CV experiments on 1c based on Savéant's work on outer-sphere (redox) catalysis 43,44 . These results are detailed in Supplementary Fig. 3 and Supplementary Table 1.
Interestingly, outer-sphere redox mediation was only observed in the presence of Li + . The redox behaviours of 1c in Li + -free TBAPF 6 and Li + -containing TBAPF 6 electrolytes are shown in Supplementary Fig. 4. In Ar-saturated environments, reversible R + /R • reduction is seen at 2.80 V in both electrolytes, so Li + does not influence R + /R • reduction. However, in O 2 -saturated, Li + -free electrolyte, no catalytic enhancement is observed. Instead, separate R + /R • and O 2 /O 2 − redox waves are present at 2.80 and 2.40 V, respectively, suggesting that the second step in the outer-sphere pathway, homogeneous electron transfer between R • and O 2 , is Li + -coupled. This type of ion-coupled catalysis, commonly seen in many proton-coupled processes 45 , is sparsely reported in the Li-O 2 field 46 . We are currently investigating the role of Li + in catalysis, but we expect that large Li + concentrations generate the driving force for Li + to pre-associate O 2 and form an Li + ···O 2 adduct that participates in the outer-sphere redox mediation.
To better understand the outer-sphere redox mediation, rate constants associated with the electrochemical and chemical steps were estimated. The experimental data were simulated using the following electrochemical and chemical steps (Supplementary Section 1 provides more information): Where k 1 is the heterogeneous electron transfer rate constant which ranged from 0.0098 to 0.012 cm s −1 , confirming fast electron transfer between the electrode and redox mediators. The resulting bimolecular rate constants for reaction between R • and Li + ···O 2 (k 2 ) are plotted with respect to E R + /R • in Supplementary Fig. 8. Noticeably, the level of catalytic enhancement increases as E R + /R • becomes more , 1b (f), 1d (g) and 4 (h). The CVs were recorded in TEGDME containing 1 M lithium triflate (LiOTF) and 3 mM redox mediator at a scan rate of 100 mV s −1 .
Article https://doi.org/10.1038/s41557-023-01268-0 negative (Fig. 3). The trend in log(k 2 ) versus E R + /R • is consistent with the model for redox catalysis rate-controlled by the electron-transfer step (activation control) 43,44 reported by Savéant, so our data in Supplementary Fig. 8 were fit to the following expression: where k s is the solution electron exchange rate constant, α is 0.5 and The fit provided a k s value of 1.774 × 10 7 M −1 s −1 , which agrees well with other studies on outer-sphere redox mediators 43,44 .
Two key conclusions were drawn from our CV analysis. First, the inner-sphere mechanism did not result in catalytic ORR for any R + derivatives investigated in this Article. Although some radicals react with O 2 , the corresponding peroxides were inert toward Li + ions, preventing closure of the catalytic cycle. Although no R + derivatives showed inner-sphere catalysis, such a process can be achieved in other derivatives whose energetics are tuned to satisfy the energy landscape in Fig. 2b. Second, cations with E R + /R • in the 2.54-2.95 V range showed detectable catalytic enhancements via the outer-sphere mechanism, but only in the presence of Li + , suggesting Li + -coupled catalysis. Catalytic rate constants increased with thermodynamic driving force for electron transfer, as predicted by the Marcus electron transfer model 43,44 .

Battery discharge
The influence of outer-sphere redox mediation on Li-O 2 discharge was evaluated as described in the Methods using either LiFePO 4 (LFP) or unprotected Li anodes. In the absence of redox mediators (Fig. 4, grey traces), the discharge quickly terminated at the 2.0 V cutoff potential after a capacity of 0.065 mAh cm −2 because of the surface-mediated O 2 reduction and carbon electrode passivation by a Li 2 O 2 film 3 . Consistent with the redox mediation observed in CVs, the discharge capacity increased in the presence of redox mediators for both LFP (Fig. 4) and Li ( Supplementary Fig. 9) anodes. However, the discharge capacities are larger for each redox mediator in LFP-containing cells, providing evidence that the reaction between R + and Li was consuming R + and effectively lowered their concentration. To further confirm the effect of outer-sphere redox mediation, differential electrochemical mass spectrometry (DEMS) measurements were carried out on Li-O 2 cells lacking redox mediators and containing 1c. Average e − /O 2 ratios of 2.53 and 2.16 were determined for the baseline and redox-mediated cells, respectively, suggesting that the predominant product in the redox-mediated cell was Li 2 O 2 .
Generally, the discharge capacities increased with R + concentration (Supplementary Figs. 10 and 11), probably due to the presence of two competing Li 2 O 2 formation pathways: (1) a surface-mediated pathway, which results in passivating Li 2 O 2 films on the electrode surface, and (2) the desired redox-mediated pathway, which moves Li 2 O 2 formation into the solution phase. At higher R + concentrations, the redox-mediated process dominates, resulting in larger discharge capacities. Given these results, the performance of each mediator was evaluated at its solubility limit (Fig. 4a). The identity and morphology of the discharge product was determined using Raman spectroscopy, scanning electron microscopy (SEM) and UV-vis titration with TiOSO 4 . Raman bands present at ~250 and 790 cm −1 (Supplementary Fig. 14) were attributed to the presence of Li 2 O 2 as a discharge product, which was further confirmed by colorimetric titration with TiOSO 4 (Supplementary Table 5). The formed Li 2 O 2 particles appear in SEM images as small flakes on the electrode surface ( Supplementary Fig. 15), matching what has been observed previously for some quinone-based redox mediators 8 .
Discharge plateaus of redox mediator-containing cells appear with lower overpotentials and tend to scale with E R + /R •. They also achieved substantially larger discharge capacities, ranging from 6-to 35-fold, which was attributed to redox mediation moving the O 2 /LiO 2 reduction into solution and mitigating surface-mediated reduction. The best performance was observed with 1b, which provided capacities competitive with those reported for quinone-based and other mediators [6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22] (DBBQ 9 is shown in Fig. 4a). Outer-sphere redox mediation often generates reactive oxygen species that have detrimental effects on cell performance; however, this is not a concern in our work. First, Li + -coupled oxygen reduction avoids O 2 − formation and instead generates LiO 2 intermediates that are non-parasitic in Li-O 2 cells 47 . Second, DEMS measurements suggest there is nearly no side-product formation during Li-O 2 cell discharge. These results provide encouraging evidence that the organic cations in Fig. 1 provide an important expansion of the library of functional redox mediators for Li-O 2 batteries.
A more quantitative comparison was made by testing the redox mediators at uniform concentrations of 25 mM (Fig. 4b). Surprisingly, the battery discharge capacity increased with more positive E R + /R •, counter to the observed relationship between E R + /R • and log(k 2 ) in the CV section. Equation (3) predicts that redox mediation rates increase as the thermodynamic driving force for electron transfer between R • and Li + ···O 2 increases, that is, the more negative E R + /R •, as expected for the normal Marcus region. Here, the opposite behaviour is observed. Results from our CV and chronopotentiometry measurements are compared in Fig. 5a, which shows how bimolecular rate constants k 2 (blue points) and battery discharge capacities (red bars) change as a function of E R + /R •. Clearly, redox mediators which react more slowly a b  with Li + ···O 2 , yield larger discharge capacities. CV measurements probe the electrochemical processes that occur on short timescales, varying in the microseconds-to-seconds range, depending on sweep rate. As such, CV is an excellent tool to study electron-transfer kinetics, but battery capacity depends on chemical processes, such as Li 2 O 2 crystal growth, which take place at timescales longer than the CV temporal window. The opposing trends displayed in Fig. 5a point to the dangers associated with relying solely on CV to screen for efficient redox mediators.
Notably, DBBQ diverges from our observed trends. DBBQ proceeds through an inner-sphere mechanism, forming a Li···DBBQ···O 2 adduct as a key intermediate, and imparting comparatively smaller bimolecular ORR rate constants ( Supplementary Fig. 16) 9,46,48 . However, the discharge capacities of DBBQ-containing cells (1.33 mAh cm −2 ) were larger than any of our redox mediators at similar concentrations ( Supplementary Fig. 16). The inner-sphere pathway must increase the reduced oxygen intermediate lifetimes (for example, Li···DBBQ···O 2 ), giving more time for diffusion away from the electrode 15,49 . Although this is beneficial, we point out that our cationic outer-sphere redox mediators can be prepared at much larger concentrations than the quinone derivatives.
The trend observed in Fig. 5a can be explained by the competition between surface-and solution-mediated O 2 /Li 2 O 2 reduction processes. The surface-mediated process favours the formation of surface-adsorbed LiO 2 intermediates, which give way to insulating Li 2 O 2 films and result in rapid cell termination 3 . The solution-mediated process favours soluble LiO 2 intermediates, which extend discharge capacities. The cell potential has a drastic effect on the competition between these processes, with the surface-mediated process being favoured at larger overpotentials 50 . Because the surface-mediated process is favoured at potentials near 2.5 V, as E R + /R • values get nearer that potential, the redox mediator becomes less able to suppress the surface-mediated process. Such mediators could be made more competitive if their concentration is substantially higher than Li + ···O 2 . This avenue of research should be explored further. At the millimolar R + concentrations used in this study, both Li + ···O 2 and R + species are available near the electrode surface, leading to competition between surface-and redox-mediated processes.
Given the relationship between discharge capacity and increasing E R + /R • values, it is probable that further improvements are expected if redox mediators with more positive potentials are utilized. However, electron transfer between R • and Li + ···O 2 will become increasingly unfavourable as E R + /R • increases, eventually leading to loss of activity. Increasing redox mediator concentration beyond standard conditions can overcome this: at high concentrations of electrochemically generated R • , electron transfer to Li + ···O 2 will become thermodynamically favourable, even for highly positive E R + /R • values, because of the equilibrium shift caused by Le Chatelier's principle. In other words, the reaction between R • and O 2 can be made thermodynamically favourable even for E R + /R • values above 2.96 V, as long as the R • concentration is much higher than the R + concentration. High R • concentrations in discharge experiments can be achieved by 'sacrificing' a fraction of current density for the conversion of R + to R • . In other words, such a process is not truly 'catalytic', because a large portion of R • is never converted back to R + . Instead, the excess R • serves to generate the requisite driving force for electron transfer from R • to O 2 . Thus, a critical question remains: how positive can E R + /R • be pushed and it still be able to mediate O 2 reduction? To determine this, the chronopotentiometric response was modelled based on redox-mediated O 2 reduction and is presented in Fig. 5b. Full discussion of the modelling can be found in the Methods and Supplementary Section 3. Shaded regions in Fig. 5b show the range of current densities (i 0 ) and R + concentrations (C 0 R + ) expected to yield effective redox mediation at specified E R + /R • (2.7, 3.0 and 3.3 V). This model uncovers key guidelines for the discovery of future redox mediators: redox mediation can be achieved for a wide range of E R + /R •, even values of 3.0 V and above, as long as the R + concentrations can be increased enough to accommodate the chosen current density.

Conclusions
An unexplored class of triarylmethyl cation discharge redox mediators for Li-O 2 batteries has been investigated using computational and experimental approaches. Inner-and outer-sphere mechanisms were explored computationally, but experimental work shows only the outer-sphere mechanism results in closure of the catalytic cycle. The increased rate of reaction between R + and O 2 with more negative E R + /R • values is indicative of outer-sphere redox mediation and Marcus-type rate dependence on the thermodynamic driving force. Enhancements to Li-O 2 discharge capacity rivalling those of well-studied quinone-based mediators were realized, revealing a class of molecular motifs whose derivatives can be explored as future redox mediators. Comparative CV and chronopotentiometric studies resulted in an interesting finding: improved discharge capacities were obtained for kinetically 'sluggish' redox mediators. More specifically, a clear   determines the minimum R + concentration required for redox catalysis at a specified E R + /R •. Points A and C illustrate that, as E R + /R • increases from 2.7 to 3.3 V, increasing R + concentrations (0.7 to 187 mM) are needed to maintain redox mediation conditions. Article https://doi.org/10.1038/s41557-023-01268-0 relationship between discharge capacity and more positive E R + /R • values was observed, which underscores the importance of controlling discharge potential to improve capacity. The observed behaviour is explained by competition between surface-mediated and redox-mediated oxygen reduction, with the surface-mediated process being less efficient as the cell potential becomes more positive. Chronopotentiometry simulations revealed that redox mediation can be achieved for E R + /R • values as positive as 2.96 V, but they must be operated at high concentrations to enable spontaneous electron transfer to O 2 . Based on the desire to reduce the discharge overpotential and prolong battery discharge times, this finding will be instrumental for the discovery of future redox mediators with more positive reduction potentials that may be prepared at high concentrations.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41557-023-01268-0.

Phenylmagnesium bromide synthesis
Phenylmagnesium bromide (Grignard) reagents were made on demand for specific redox mediator synthesis. In a dry flask attached to a Schlenk line, with a dry stir bar, 1 equiv. of dried Mg powder and a few small crystals of I 2 were added and the headspace was purged with Ar gas. One equivalent of bromobenzene dissolved in dry tetrahydrofuran (THF; 20 ml) was added to the Mg powder. The solution was briefly heated to begin the reaction and then placed in an ambient-temperature water bath. The reaction mixture was then stirred for 60 min. Once completed, the crude Grignard reagent was used without isolation for subsequent steps.

9-phenyl-10-methylacridinium triflate (1a)
1a was made by a modified procedure taken from the literature 36 . First the acridinol analogue (9-phenyl-10-methylacridin-9-ol) was made by first dissolving 10-methylacridinone (0.50 g, 2.39 mmol, 1 equiv.) in dry THF (50 ml), then the 10-methylacridinone was added dropwise to an Ar-purged, phenylmagnesium bromide-containing (2.2 equiv.) Schlenk flask over a 20-min period. The flask was chilled to 0 °C in an ice bath during 10-methylacridinone addition, then allowed to warm to room temperature and stirred for 24 h. The reaction was quenched with aqueous NH 4 Cl (10 ml) and diluted with ethyl acetate (EtOAc) and H 2 O. The aqueous layer was extracted with EtOAc three times. The resulting organic layer was dried over anhydrous MgSO 4 . The solvent was evaporated, yielding a solid white product that matched the published spectrum 36 .
To make 1a, 9-phenyl-10-methylacridin-9-ol (0.25 g, 0.87 mmol) and a magnetic stir bar were added to a dry 15-ml vial that was purged with Ar. The 9-phenyl-10-methylacridin-9-ol was dissolved under stirring in dry Et 2 O at room temperature. Triflic acid (0.082 ml, 0.97 mmol, 1.1 equiv.) was added to the vial and the mixture was stirred for an additional 30 min and then chilled to 0 °C. The precipitate was filtered, washed with anhydrous Et 2 O and dried in vacuum to give a solid product. The 1 H NMR spectrum of 1a matched the previously published spectrum 36 .

9-(2,6-dimethoxyphenyl)-1,8-dimethoxy-10-methyacridinium perchlorate (1d)
1d was synthesized by a modified procedure from the literature 38 . First, the precursor tris(2,6-dimethoxyphenyl)methylium perchlorate was made. 1,3-Dimethoxybenzene (2 ml, 15.4 mmol, 3.5 equiv.) was dissolved in dry THF (10 ml) under argon, then 2.5 M n-butyl lithium in hexane (6.2 ml, 15.4 mmol, 3.5 equiv.) was added dropwise at −78 °C. After 30 min, the solution was warmed to room temperature and stirred for 1 h. Diethyl carbonate (0.52 ml, 4.4 mmol, 1 equiv.), dissolved in 25 ml of benzene, was added to the solution, which was stirred and refluxed for two days. The reaction was quenched in water and extracted with dichloromethane. Combined CH 2 Cl 2 layers were added, dried, and the solvent was evaporated, yielding a brownish residue. The product was dissolved in acetonitrile, perchloric acid was added, and a precipitate formed. The precipitate was collected, washed with Et 2 O, and dried under vacuum. The 1 H NMR spectrum matched the published spectrum 38 .
To make 1d, tris(2,6-dimethoxyphenyl)methylium perchlorate (1 g, 2 mmol) was dissolved in 15 ml of N-methylpyrrolidine to which 1.2 ml of methylamine (33% in ethanol, 9.6 mmol, 4.8 equiv.) was added. The mixture was stirred for 12 h, then poured into Et 2 O under stirring. The red precipitate was collected, washed with Et 2 O and dried under vacuum. The 1 H NMR spectrum of 1d matched the published spectrum 38 .

9-phenylxanthylium triflate (2a)
2a was synthesized by a modified procedure from the literature 36 . First, the xanthenol analogue (9-phenyl-xanthen-9-ol) was made as follows: xanthone (0.50 g, 2.54 mmol, 1 equiv.) was dissolved in dry THF (50 ml), then the xanthone was added dropwise to an Ar-purged, phenylmagnesium bromide-containing (2.2 equiv.) Schlenk flask over a 20-min period. The flask was chilled to 0 °C in an ice bath during xanthone addition, then allowed to warm to room temperature and stirred for 24 h. The reaction was quenched with aqueous NH 4 Cl (10 ml) and diluted with EtOAc and H 2 O. The aqueous layer was extracted with EtOAc three times. The resulting organic layer was dried over anhydrous MgSO 4 . The solvent was evaporated, yielding a solid white product that matched the published spectrum 36 .
To make 2a, 9-phenyl-xanthen-9-ol (0.25 g, 0.91 mmol, 1 equiv.) and a magnetic stir bar were added to a dry 15-ml vial that was purged with Ar, then the 9-phenyl-xanthen-9-ol was dissolved under stirring in dry Et 2 O at room temperature. Triflic acid (0.085 ml, 1.00 mmol, 1.1 equiv.) was added to the vial, and the mixture was stirred for an additional 30 min, then chilled to 0 °C. The precipitate was filtered, washed with anhydrous Et 2 O and dried in vacuum to give solid product. The 1 H NMR spectrum of 2a matched the previously published spectrum 36 .

9-(4-tolyl)xanthylium perchlorate (2b)
2b was synthesized by a modified procedure from the literature 37 . First, the xanthenol analogue, 9-(4-tolyl)-9H-xanthen-9-ol was made as follows: 1,3,5-trimethylbenzene (72 mg, 0.6 mmol) was dissolved in 1.3 ml of dry THF, and the solution was purged with Ar, then n-butyl lithium (0.6 mmol, 0.24 ml, 2.5 M) was added to the mixture dropwise at −78 °C. The resultant mixture was warmed to room temperature and stirred under Ar for 2 h. The mixture was cooled to 0 °C and xanthone (0.1167 g, 0.6 mmol) dissolved in a minimal amount of toluene was added to the solution. The mixture was stirred for 24 h and allowed to warm to room temperature during that time. The reaction was quenched with water and the product was extracted with dichloromethane. Combined CH 2 Cl 2 layers were added and dried over anhydrous MgSO 4 . The solvent was evaporated and yielded a white solid product. The 1 H NMR spectrum of 9-(4-tolyl)-9H-xanthen-9-ol matched the already published spectrum 37 .
To make 2b, a 15-ml vial was charged with 9-(4-tolyl)-9Hxanthen-9-ol (0.50 mmol) and concentrated perchloric acid (70%, 4 ml). The resulting mixture was stirred overnight. The precipitate was filtered off, washed with anhydrous diethyl ether (Et 2 O, 15 ml), and dried in vacuum to give a solid product. The 1 H NMR spectrum matched the previously published spectrum 37 .

9-(2,4,6-trimethoxyphenyl)xanthylium perchlorate (2c)
2c was synthesized by a modified procedure from the literature 37 . First, the xanthenol analogue, 9-(2,4,6-trimethoxy)-9H-xanthen-9-ol was made as follows: 1,3,5-trimethoxybenzene (100 mg, 0.6 mmol) was dissolved in 1.3 ml of dry THF and the solution was purged with Ar, then n-butyl lithium (0.6 mmol, 0.24 ml, 2.5 M) was added to the reaction mixture dropwise via canula at −78 °C. The reaction mixture was allowed to warm to room temperature and stirred under Ar for 2 h. The mixture was cooled to 0 °C and xanthone (0.1167 g, 0.6 mmol) dissolved in a minimal amount of toluene was added to the solution. The mixture was stirred for 24 h and allowed to warm to room temperature during that time. The reaction was quenched with water followed by extraction with CH 2 Cl 2 . Organic layers were merged and dried over anhydrous magnesium sulfate. After the CH 2 Cl 2 was evaporated, it yielded a white Nature Chemistry Article https://doi.org/10.1038/s41557-023-01268-0 solid. The 1 H NMR spectrum of 9-(2,4,6-trimethoxy)-9H-xanthen-9-ol matched the already published spectrum 37 .

4,8,12-trimethyl-4,8,12-triazatriangulenium perchlorate (5)
5 was synthesized by a modified procedure from the literature 51 . 4 (100 mg, 0.24 mmol) was dissolved in 10 ml of N-methylpyrrolidine, then 1.2 ml of methylamine solution in ethanol (33%, 6.2 mmol, 26 equiv.) was added and the mixture was stirred at 90 °C for 8 h under ambient conditions. The reaction was allowed to cool and was poured into a cooled 50% aqueous solution of HClO 4 . The precipitate was filtered, washed with Et 2 O and dried under vacuum. The 1 H NMR spectrum of 5 matched the published spectrum 51 .
Chemically derived peroxide 9-(4-tolyl)xanthyl peroxide 2b (2b/ RO-OR) was prepared according to published procedures 37 and its putative structure is reported in Supplementary Section 1. 1  Electrochemically derived peroxide derivatives (2a/RO-OR and 2b/RO-OR) were prepared and collected through amperostatic coulometry. In O 2 -saturated solutions of 2a and 2b, a constant current (50 μA) was applied to the cell for 10-s intervals, and the solid was collected from the glassy carbon (GC) electrode and dissolved in THF. The electrochemically derived 2a/RO-OR and 2b/RO-OR showed the same NMR peaks as their chemically derived counterparts.

Electrochemistry
Tetrabutylammonium perchlorate was recrystallized from methanol. Lithium triflate (LiOTF) was dried under vacuum at 80 °C for three days before use. TEGDME was purchased from Millipore-Sigma and dried over 3-Å activated molecular sieves before use. The final water content of TEGDME was <100 ppm as measured by Karl Fischer titration. CV measurements were performed with a standard three-electrode set up with GC, 10 mM Ag/AgNO 3 and a Pt wire serving as the working, reference and counter electrodes, respectively. A Gamry 1010B potentiostat was used for CV measurements.

CV
All CV measurements were performed in a three-electrode cell set-up using GC (MF-2012, Bioanalytical systems, 0.075-cm 2 surface area), 10 mM Ag/AgNO 3 non-aqueous reference electrode (MF-2062, Bioanalytical systems) and Pt wire counter electrode (MW-4130 or MW-1033, Bioanalytical systems) and a Gamry 1010B potentiostat. Working electrodes were polished before each individual CV using alumina (CF-1050, Bioanalytical systems) and diamond (MF-2054, Bioanalytical systems) slurry polishes. CV experiments were conducted in dry TEGDME containing 1 M LiOTF and 3 mM redox mediator. For O 2 -free and O 2 -saturated measurements, solutions were purged for 20 min with Ar or O 2 gas, respectively. [O 2 ] in CV solutions was measured using a Neo-FoxGT O 2 sensor provided by Ocean Insight.

Battery testing
Swagelok Li-O 2 cells were assembled in an Ar-filled (MBraun) glovebox with H 2 O and O 2 levels below 0.1 and 3 ppm, respectively, similar to previous descriptions 9 . Carbon gas diffusion layers (GDLs) were Freudenberg H23C2 GDLs and were provided by the Fuel Cell Store. For moisture removal, GDLs were dried under vacuum at 150 °C for three days before storage in a glovebox. During assembly, an LFP (3.5 V) 52 or Li metal anode was separated from a single, 5.6-mm diameter GDL disk by a Whatman glass-fibre paper separator. Redox mediator-containing, 1 M LiOTF TEGDME electrolyte (~0.150 ml) was added to the cathode side, and a stainless-steel-mesh current collector was used to cover the GDL. The assembled cells were placed inside hermetically sealed glass tubes, purged with dry O 2 and discharged on a Maccor battery cycler.
After discharge, the Swagelok Li-O 2 cells were disassembled to recover the carbon GDLs. The GDLs were washed with a small amount of dimethoxyethane and dried on a hot plate in the glovebox. After drying, the GDLs, deposits and films were characterized by SEM and Raman spectroscopy.

DEMS
DEMS measurements were performed according to a previously established method 47,53 . In short, a DEMS cell (MTI Corp) was assembled, using the same LFP anode, glass-fibre separator and carbon GDL as used in Swagelok-cell testing, inside the glovebox, and connected to the DEMS device. Once connected, the DEMS cell was purged with O 2 gas to ~20 p.s.i. and allowed to rest at open-circuit potential for 1 h. The overall cell volume was calibrated using a previously specified volume exchange technique 47,53 . The assembled cell volume was calibrated by replacement with five different standard volume loops (50, 100, 250, 500 and 1,000 μl) and comparing the partial pressure change within the cell. The overall volume was determined to be 6 ml. During discharge, the pressure inside the cell volume, and thus consumption of O 2 , was monitored by an in-line pressure transducer. Oxygen consumption was compared to the discharge current to determine the mol e − consumed per mol O 2 .

Li 2 O 2 quantification
Li 2 O 2 quantification was performed using a previously reported UV-vis method 9,54 . In brief, the Li-O 2 cell was deconstructed and a glass-fibre separator and GDL were added to a small vial with 5 ml of TiOSO 4 in aqueous H 2 SO 4 . Li 2 O 2 is known to spontaneously react with water to form H 2 O 2 and, in the presence of TiOSO 4 , a yellowish [Ti(O 2 )] 2+ complex is formed with absorption at 410 nm. A small quantity of each solution was tenfold diluted to generate appropriate UV-vis solutions, which were compared with a calibration curve made with commercial Li 2 O 2 (Aldrich).

Raman spectroscopy
Raman spectroscopy of the GDLs and their discharge products was performed with a Renishaw inVia Reflex microscope using either 633-nm (red) or 532-nm (green) laser light at the Electron Microscopy Core (EMC) at UIC's Research Resources Center (RRC). Inside the glovebox, the GDLs were loaded into an air-sensitive Raman cell with a thin quartz glass window for observation.

SEM imaging
SEM characterization was carried out using a JEOL JSM-IT500HR SEM in field-emission mode that is part of the EMC at the UIC RRC. Cleaned and dried GDLs were removed from the Ar glovebox in tightly sealed vials https://doi.org/10.1038/s41557-023-01268-0 and transferred as quickly as possible to the SEM vacuum chamber to minimize air exposure. All images were taken at ×6,000 magnification using a 5-keV acceleration voltage.

DFT calculations
All calculations were performed on Gaussian 09 or 16 55 software using the computational resources from the Laboratory Computing Resource Center at Argonne National Laboratory. Gas-phase optimization was performed using the B3LYP or uB3LYP hybrid functional 56,57 and 6-311g(d) 58 basis set. Single-point energies were calculated at the same level of theory using Et 2 O solvation as implemented in the IEFPCM model 59 . Vibrational frequency analysis showed the presence of no imaginary frequencies in our optimized structures. Detailed information on the evaluation of each step in the catalytic mechanism is provided in Supplementary Section 2.

Chronopotentiometry modelling
Chronopotentiometry simulations were performed using a model involving electron transfer and a follow-up chemical step, as defined by equations (1) and (2), which has been defined previously 60 . Full details on the modelling are available in Supplementary Section 3.

Data availability
Source data for figures in the main text have been uploaded to the Figshare public data repository and can be accessed at https://doi. org/10.6084/m9.figshare.21719882 (ref. 61). Source data are provided with this paper.